Expected value is the most useful number in sports betting — and one of the most misunderstood. Here is what it actually means, how it is calculated, and why it is the right thing to target instead of win rate.
The basic idea
EV is the average outcome per dollar risked, over many bets.
Every bet you place has an implied outcome: if you made the exact same bet hundreds of times, what would your average return be? That average is expected value.
A positive EV bet means the average outcome is in your favor. A negative EV bet means the sportsbook has the edge. Most bets at standard juice (-110) are slightly negative EV for the bettor — the book's margin is built into the price.
Winning a bet does not mean it was positive EV. Losing a bet does not mean it was negative EV. EV describes the long-run expectation, not any single result.
The formula
Win probability × payout minus loss probability × stake.
EV = (p_win × payout) − (p_loss × 1)
At standard juice (-110), a win pays 0.909 units. So a bet with a 55% win probability has EV of (0.55 × 0.909) − (0.45 × 1) = 0.500 − 0.450 = +0.050, or about +5% per unit risked.
The same bet at 50% win probability has EV of (0.50 × 0.909) − (0.50 × 1) = −0.046. Negative. The difference between 50% and 55% is the difference between giving money to the book and taking it.
Where win probability comes from
This is the part most betting tools skip over.
EV is only as good as the win probability estimate behind it. If the probability is wrong, the EV number is wrong — even if the math is correct.
Model-derived
A statistical model trained on historical game data estimates the probability of covering or hitting a total based on team form, rest, pace, and other features.
Market-implied
The sportsbook line itself encodes a probability. A -110 spread implies the market gives each side roughly equal chance of covering. Large line moves signal updated market opinion.
Calibrated
A well-calibrated model is one where 60% predictions win about 60% of the time. Calibration matters because an overconfident model produces inflated EV estimates that don't materialize.
Why win rate is not the right target
A 60% win rate at bad prices loses money. A 52% win rate at good prices makes it.
Win rate is outcome-focused. EV is process-focused. A tout who posts only huge favorites will show a high win rate and still lose money long-term because the payouts are too small to compensate for the losses.
The right question is not "did this bet win?" It is "did this bet have positive expected value at the price I took?" You will not know if a single bet was correctly priced. But if you consistently find positive EV, the long-run math works in your favor.
EV and price
The same opinion has different EV at different prices.
If you think a team has a 55% chance to cover, a -110 line gives you +5% EV. The same bet at -120 gives you roughly +0.4% EV. At -130, it is negative EV — you are paying too much for the same opinion.
This is why line shopping matters. Getting -108 instead of -115 on a spread bet is not a minor convenience. It is the difference between a positive-EV bet and a break-even one over hundreds of similar plays.
How Linelabs uses EV
EV is one filter, not the whole story.
The live board shows an EV estimate for each surfaced play. It is derived from a calibrated XGBoost model trained on historical NBA game data, compared against the current line at the best available book. Plays below a minimum EV threshold are not surfaced.
But EV is one signal, not a guarantee. The board also shows win probability, confidence tier, current price, and freshness — because a positive EV estimate from this morning is less useful if the line has moved significantly since. Price context is shown alongside EV so you can decide whether the current number still makes the play actionable.
Linelabs does not claim that surfaced plays are guaranteed to win or that the EV estimates are precisely correct. They represent a model's best current read. Sample size, calibration drift, and line movement all affect whether historical EV translates to future outcomes.